Deciphering Selectivity in Organic Reactions: A Multifaceted Problem

Article pubs.acs.org/accounts

Deciphering Selectivity in Organic Reactions: A Multifaceted Problem Published as part of the Accounts of Chemical Research special issue “Computational Catalysis for Organic Synthesis”. David Balcells,† Eric Clot,‡ Odile Eisenstein,*,†,‡ Ainara Nova,† and Lionel Perrin§ †

Centre for Theoretical and Computational Chemistry (CTCC) and The Department of Chemistry, University of Oslo, P.O. Box 1033, Blindern, 0315 Oslo, Norway ‡ Institut Charles Gerhardt, UMR 5253 CNRS-UM-ENSCM, Université de Montpellier, Place Eugène Bataillon, 34095 Cedex 5 Montpellier, France § Univ Lyon, Université Lyon1, CNRS, INSA, CPE-Lyon, ICBMS, UMR 5246, 43, Bd du 11 Novembre 1918, 69622 Cedex Villeurbanne, France CONSPECTUS: Computational chemistry has made a sustained contribution to the understanding of chemical reactions. In earlier times, half a century ago, the goal was to distinguish allowed from forbidden reactions (e.g., Woodward−Hoffmann rules), that is, reactions with low or high to very high activation barriers. A great achievement of computational chemistry was also to contribute to the determination of structures with the bonus of proposing a rationalization (e.g., anomeric effect, isolobal analogy, Gillespie valence shell pair electron repulsion rules and counter examples, Wade−Mingos rules for molecular clusters). With the development of new methods and the constant increase in computing power, computational chemists move to more challenging problems, close to the daily concerns of the experimental chemists, in determining the factors that make a reaction both efficient and selective: a key issue in organic synthesis. For this purpose, experimental chemists use advanced synthetic and analytical techniques to which computational chemists added other ways of determining reaction pathways. The transition states and intermediates contributing to the transformation of reactants into the desired and undesired products can now be determined, including their geometries, energies, charges, spin densities, spectroscopy properties, etc. Such studies remain challenging due to the large number of chemical species commonly present in the reactive media whose role may have to be determined. Calculating chemical systems as they are in the experiment is not always possible, bringing its own share of complexity through the large number of atoms and the associated large number of conformers to consider. Modeling the chemical species with smaller systems is an alternative that historically led to artifacts. Another important topic is the choice of the computational method. While DFT is widely used, the vast diversity of functionals available is both an opportunity and a challenge. Though chemical knowledge helps, the relevant computational method is best chosen in conjunction with the nature of the experimental systems and many studies have been concerned with this topic. We will not address this aspect but give references in the text. Usually, a computational study starts with the validation of the method by means of benchmark calculations vs accurate experimental data or state-of-the-art calculations. Finally, computational chemists can bring more than the sole determination of the reaction pathways through the analysis of the electronic structure. In our case, we have privileged the NBO analysis, which has the advantage of describing interactions on the basis of terms and concepts that are shared within the chemical community. In this Account, we have chosen to select representative reactions from our own work to highlight the diversity of situations than can be addressed nowadays. These include selective activation of C(sp3)−H bonds, selective reactions with low energy barriers, involving closed shell or radical species, the role of noncovalent interactions, and the importance of considering side reactions.

■

INTRODUCTION Chemical reactions must be efficient and selective to be usable in synthesis. Computations can now contribute to improve chemical transformations by determining the factors that control the efficiency and the selectivity through the determination of reaction pathways. In this Account, we provide selected examples of the factors that were identified. We consider reactions where selective C(sp3)−H bond activation is achieved through combining agostic interactions involving the unactivated C−H bond and the assistance of a coordinated base. We mention the importance of understanding the pathways leading to side products or catalyst deactivation for improving both robustness © XXXX American Chemical Society

and selectivity. Several examples are devoted to reactions that have low activation barriers but are selective. Some of them involve radical reactions whereas others occur between closedshell systems. Steric bulk tends to disfavor reactions, but we give a counterexample where it promotes product formation. In many examples, noncovalent interactions appropriately placed can induce high selectivity. We conclude by giving an example based on dynamic kinetic resolution in which high selectivity originates from a complex network of equilibria all leading to the same exit Received: February 24, 2016

A

DOI: 10.1021/acs.accounts.6b00099 Acc. Chem. Res. XXXX, XXX, XXX−XXX

Article

Accounts of Chemical Research channel yielding the desired product with high selectivity. We do not address here the problems that arise during a computational study like appropriate modeling and methods. Some of these issues were described in recent articles.1−5

metalation to the coordinating group. Another alternative to trigger the formation of a M−C bond for functionalization at a specific position is to anchor the substrate through Pd(0)catalyzed C−X oxidative addition (X = halide). This was shown by Fagnou, among others, to efficiently yield biaryl via direct arylation (Scheme 1b).9 In these reactions, computations have identified the critical role played by the base (acetate, carbonate, pivalate, etc.) to induce simultaneous Pd−C bond formation and C−H bond breaking once coordinated to the Pd(II) center through concerted metalation deprotonation (CMD) independently known as ambiphilic metal−ligand activation (AMLA) mechanisms.10 Initially, many developments concerned the Pdcatalyzed formation of C(sp2)−C(sp2) bonds via direct arylation. In 2003, Baudoin et al. disclosed one of the first examples of Pdcatalyzed C(sp3)−H bond activation (Scheme 1c).11 The presence of both P(o-Tol)3 as ligand to Pd and a base (carbonate) is essential to maximize the product of C(sp3)−H activation (2) with respect to the product of dehalogenation (3). Further studies showed that the nature of the substituents on the benzylic carbon has a great impact on the products (Scheme 1d,e). A detailed combined experimental/computational study of the mechanism of formation of benzocyclobutene by Pd-catalyzed C−H activation was carried out.12 Optimization of the catalytic conditions resulted in the choice of Pd(OAc)2 (10 mol %) as the palladium source, P(t-Bu)3 as the phosphine ligand (20 mol %), K2CO3 (1.3 equiv) as the base in DMF at 140 °C. DFT calculations on the mechanism of formation of benzocyclobutene from 2-bromo-tert-butylbenzene catalyzed by Pd−P(t-Bu)3 (Scheme 1e) indicated that C−Br oxidative addition to a monoligated Pd(0) complex has low energy barrier. In addition, depending on the relative position of the phosphine ligand and the metalated aromatic ring, two different complexes can be obtained upon substitution of bromide by the base (Figure 1).

■

SELECTIVE C−H BOND ACTIVATION AND SPECTATOR AGOSTIC INTERACTION Transition-metal catalyzed C−H functionalization is a powerful tool to transform otherwise unreactive ubiquitous C−H bonds into value-added targets through an atom-economy transformation.6,7 However, due to both the large number of different C−H bonds in a molecule and the weak Lewis-basic character of these bonds, activity and selectivity are two important aspects to consider in C−H bond activation and subsequent functionalization. One breakthrough was disclosed by Murai with the chelation-assisted hydroarylation of ortho-C−H bonds (Scheme 1a).8 In this transformation, an electron-donating group (EDG = ketone, imine) substituting an aromatic ring coordinates to the metal allowing a specific agostic interaction to develop between the ortho-C−H bond and the metal, activating the former. Further evolution of the system, either by C−H oxidative addition or σ-CAM transformation triggers selective ortho Scheme 1. Selective C−H Bond Activation

Figure 1. Three-dimensional representation of cis and trans precursor complexes and subsequent AMLA-transition states for C−H(t-Bu)activation in [Pd(P(t-Bu)3)(CO3)(2-t-Bu-C6H5)]−.

From the κ2-complex cis, C−H activation is effective through a typical AMLA-like transition state (TScis, Figure 1) with an energy barrier of ΔE# = 45.0 kcal/mol, which is higher than that from the other isomer (TStrans, ΔE# = 27.6 kcal/mol above trans). This was ascribed to several factors. The κ1-coordination of the base imposed by the geometry of trans precludes any B


Article

Accounts of Chemical Research

Figure 2. Pathways for 1,2-palladation and 1,4-palladium migration.

Figure 3. Competitive pathways (ΔG in kcal/mol) for benzocyclobutene and olefin formation.

extra-bonding, as in the κ2-adduct, which would need to be lost in the TS. The formation of an agostic interaction in trans increases the acidity of the geminal hydrogen that is abstracted by the base in TStrans. After dissociation of the protonated base, the C−C coupling is effective (ΔE# = 22.5 kcal/mol) regenerating the monoligated Pd−P(t-Bu)3 catalyst. The calculations are in agreement with the observations pointing to a C−H activation rate-determining step. In addition, the computational study identified an original mechanism associated with 1,4-Pd migration observed with certain fluorinated substrates and by deuterium labeling studies. In this mechanism, the position of the base in TStrans is essential to open a pathway resulting in a formal 1,4-Pd migration (Figure 2). The driving force for this migration, in the case of the fluorinated compound, is the stronger Pd−C bond ortho to a fluorine.13 Even though valuable information on reaction mechanism was obtained by calculations using model systems, such simplifications may lead to erroneous conclusions. In the case of the formation of benzocyclobutene described above, the reaction pathway was computed with P(t-Bu)3 and PMe3 as ligands and acetate, bicarbonate, and carbonate as bases to test their influence.14 Even though the κ2-coordinated base complex is significantly more stable than the κ1-coordinated system, the calculations showed that bromide substitution to access these intermediates critically depends on the nature of the phosphine and the base. With PMe3 and acetate, formation of the κ2-isomer is kinetically favored, whereas with P(t-Bu)3 and carbonate, access to the κ1-isomer is preferred. The substitution step, not

often included in the computations, has important consequences. Along the pathway from the κ2-isomer, the experimental trend observed with the different bases (rate carbonate ≫ bicarbonate = acetate) could not be reproduced. This is essentially because the base has to break the κ2-coordination along the C−H activation pathway, which is not easy for a strongly interacting base like carbonate. In addition, the abstracted hydrogen is involved in an agostic interaction at the TS (TScis Figure 1), thus reducing its protic character. In contrast, from the κ1-isomer, the action of the base is essentially to abstract a proton geminal to an agostic C−H bond. In this case, the most efficient base is carbonate as observed experimentally. This study showed that considering the actual experimental system is sometimes necessary to reach full understanding of selectivity issues. The pathway through the κ1-isomer identified an important feature: the C−H bond that is activated should be on a carbon that also bears a C−H bond ready to be engaged in an agostic interaction (primary or secondary positions). In fact, when the benzylic carbon bears only isopropyl groups, formation of benzocyclobutene is impeded and an indane is formed (Scheme 1f).15 The DFT calculations on the real system indicated that the C−H activation at the primary position is preferred over the one at the tertiary position by ca. 7.7 kcal/mol. In addition, the C−H activation leading to the major diastereoisomer is ca. 1.3 kcal/mol lower than that leading to the minor diastereoisomer. Using a chiral phosphine ligand such as binepine allowed development of an asymmetric version of this indane synthesis.16 DFT-D3 calculations reproduced both C


Article


less, the rebound mechanism is used by hydroxylases (e.g., cytochrome P450) to catalyze the C−H to C−OH transformation in a highly selective manner.20 In contrast, desaturases (e.g., stearoyl-CoA) can selectively catalyze C(H)-C(H) to C C.21 In collaboration with Crabtree, we showed that [Mn(tpp)(Cl)] (tpp = tetraphenylporphyrin) is one of the few catalysts capable of mimicking both enzymatic functions.22 Previous studies by us23 and others24 showed that this complex has singlet, triplet, and quintet spin states within a narrow range of energies, tuned by the trans ligand.25 In the open-shell states, the metal center has oxyl character, MnIV−O•, promoting radical H abstraction from the aliphatic (CH2)2 moiety of the substrate.22 The triplet oxyl state of [Mn(O)(tpp)(Cl)] yields a Mn(IV)− hydroxo complex bound to the resulting organic radical, i.e. [Mn•(OH)(tpp)(Cl)]·MHP• (MHP• = 9-monohydrophenanthrene radical) with a very low energy barrier of 3.3 kcal/mol. A key feature of this transient intermediate is the electronic structure of the metal center; the associated SOMO (singly occupied molecular orbital) has relevant contributions from both Mn(3d) and O(2p) orbitals. This yields a spin density in which the oxyl has a cylindrical shape oriented perpendicular to the MnOH plane (Figure 4). Due to the directional character of the

the enantiomeric and diastereoisomeric ratio observed experimentally. The preference for one particular enantiomer was ascribed to a more efficient network of stabilizing weak interactions within the crucial TS for C−H activation as revealed by an NCI analysis. Further study of the competition between formation of benzocyclobutene and olefin highlighted the difficulty to reproduce the experimental observations (Scheme 1g) using only computed free energy profiles.17 Both α- and β-positions of the ethyl substituent fulfill the requirement to have two C−H bonds on the same carbon to favor the C−H activation. The C− H activation at the α-position is preferred over that at the βposition by ca. 4.2 kcal/mol. Because C−C coupling to form indane is computed to be easier than C−C coupling to form benzocyclobutene, the calculations are in agreement with the nonobservation of indane derivatives. It shows that indane could be obtained only in the case of tertiary α-position. The C−H activation at the secondary α-position leads to a five-membered palladacycle (A, Figure 3) that is the crucial intermediate governing selectivity. Dissociation of the protonated base forms B, which leads to the benzocyclobutene product with ΔG# of 27.4 kcal/mol. Similarly to the mechanism found for 1,4-Pd migration (Figure 2), proton transfer from HCO3− in A to the aromatic ring generates the κ2-intermediate C. Rotation around the Pd−C bond in C allows the carbonate to abstract a proton from the methyl group, thus yielding the olefin complex D. This last C−H activation has a ΔG# of 28.6 kcal/mol. Comparison of the activation barrier forming benzocyclobutene and olefin indicates that the former should be the major species, contrary to experiment (Scheme 1g). However, the crucial step deciding the final product ratio is the competition between HCO3− dissociation from A and proton transfer to form C. Unfortunately, accurate dissociation energetics of HCO3− are difficult to compute. Therefore, a kinetic model was considered where the rate constants were extracted from the computed ΔG# and ΔG values using the Eyring equation. The only assumed value in the overall kinetic model was the activation barrier for HCO3− dissociation. The kinetic model with a ΔG# value of 6 kcal/mol resulted in the sole formation of the benzocyclobutene, whereas with a ΔG# of 12 kcal/mol the experimental ratio between benzocyclobutene and olefin could be reproduced qualitatively. This shows that, even though computations can address many aspects of selectivity in catalysis, some processes are still difficult to represent. Nevertheless, kinetic modeling based on identified transition states and intermediates associated with the computed free energy surfaces allow improved understanding.

Figure 4. Rotational switch in the C−H oxidation of dihydrophenanthrene by [Mn(tpp)(Cl)].

oxyl spin, the rotation of the Mn−OH becomes a switch between two reactions. Under rotation in one direction, the oxyl spin vector reorients toward the C radical of MHP• yielding hydroxylation by rebound, whereas, under the other direction, the oxyl spin vector reorients toward the vicinal CH2 moiety yielding dehydrogenation by H abstraction. The calculations suggested that both reactions are extremely fast with energy barriers lower than 1 kcal/mol. This avoids free diffusion of the MHP• radical, preventing the formation of homocoupling products, in agreement with experiments. In contrast to the enzymatic systems, in which this switch can be controlled by weak interactions involving the amino acid residues, the free rotation of the Mn−OH bond in the [Mn•(OH)(tpp)(Cl)]· MHP• intermediate yields the product mixtures observed in the experiments. In some catalytic systems, the final outcome of the reaction depends on the interplay between different spin states, as in the two-state reactivity model proposed by Shaik.26 DFT calculations in collaboration with Pérez27 showed that hydroxylation/ dehydrogenation selectivity in copper-catalyzed C−H oxidations can be controlled by spin crossover.28 By using hydrogen peroxide as oxidant, [Cu(TpBr3)] (TpBr3 = tribromo-trispyrazolylborate) catalyzes the hydroxylation of cyclohexane. The triplet ground state of the postulated active species, [CuIII(O)(TpBr3)], has radical oxyl character and abstracts one H from cyclohexane with a low energy barrier of 10.0 kcal/mol. The resulting [CuII(OH)(TpBr3)] recovers the closed-shell Cu(I)

■

SELECTIVITY IN HOMOLYTIC C−H BOND CLEAVAGE Under oxidative conditions, transition metal complexes with strongly donating ligands can promote C−H hydroxylation of organic substrates. This chemistry involves the rebound mechanism, in which a highly reactive metal−oxo intermediate triggers the homolytic cleavage of the C−H bond.18 The resulting OH moiety, which is stabilized by the metal center in a hydroxo intermediate, rebounds to the carbon radical yielding the alcohol product. Further, when the radical intermediate contains the •CHCH2 moiety, the hydroxo intermediate may promote dehydrogenation to an alkene by means of a second H abstraction.19 Hydroxylation and dehydrogenation can have similar thermodynamics and kinetics parameters, yielding lowselectivity reactions with complex product mixtures. NonetheD


Article


Figure 5. TS-like MECP structures associated with the hydroxylation (left) and dehydrogenation (right) of cyclohexane by [Cu(TpBr3)]. Cleaving and forming bonds are highlighted in red and blue, respectively.

■

one of the diasterotopic faces of the exocyclic double bond.30 The calculated dr ratio of 96:4 is very close to the experimental value. The analysis of the results revealed the surprising preference for a conformation of the artemisinate (dominating form at pH = 9) that was due to a network of stabilizing interactions between various C−H bonds, notably the most polar vinylic C−H bond of the internal double bond and the electron rich double bond to be reduced. The preferred conformation has a more accessible face for reaction with HNNH. These stabilizing interactions were also found into the reactant-like lowest TS, suggesting a determining contribution of the most stable conformer. These weak nonclassical attractive interactions, analyzed with NCI-plot, revealed the key role played by the vinylic CH bond (Figure 7). The preferential conformation was confirmed by an NMR study of artemisinate in the conditions of reaction. The predicted preferential conformation of artemisanate in the conditions of reaction was confirmed by NMR studies.

Figure 6. Selective formation of dihydro-artemisinic acid (a key step in the synthesis of artemisine, a powerful drug against malaria).

CONTROLLING SELECTIVITY VIA MOLECULAR RECOGNITION Molecular recognition mechanisms,31 which are ubiquitous in enzymatic catalysis, can be a powerful approach to the selective activation of inert C−H bonds in functionalized organic molecules. Nonetheless, their implementation in homogeneous molecular systems is challenged by the features required, which include reversible catalyst−reactant binding, effective functional group protection, and unrecognized reactant exclusion. One of the few successful examples of artificial molecular recognition is the dinuclear [Mn(terpy′)(H2O)(μ-O)]23+ (terpy′ = terpyridine functionalized with Kemp’s triacid) complex developed by Crabtree and Brudvig (Figure 8).32 This species catalyzes the selective C−H oxidation of ibuprofen at the secondary benzylic position, thus preventing the oxidative decarboxylation of the more reactive other benzylic position. DFT calculations on the reaction mechanism revealed a low-energy rebound pathway involving a double H-bond between the carboxylic groups of the substrate and the catalyst.33 This interaction keeps the most reactive part of ibuprofen at the recognition site meanwhile exposing the p-benzylic position to the oxyl. This interaction is preserved at the TS although the system stretches to undergo H abstraction resulting in a low energy barrier of 5.1 kcal/mol. In

catalyst by triplet-to-singlet spin crossover. Unexpectedly, the relaxation of the associated MECP (minimum energy crossing point) yields cyclohexanol upon full geometry optimization in the singlet state; hence, hydroxylation is triggered by spin crossover. The geometrical features of the MECP are indeed analogous to that of a classical rebound transition state (Figure 5), with concomitant cleavage and formation of the Cu−OH and C−OH bonds, respectively. Again, the reoptimization of the MECP from a geometry guess based on a TS for H abstraction converged into a different crossing point yields cyclohexene upon relaxation on the singlet state; hence, spin crossover can also trigger dehydrogenation, which was thereafter detected experimentally. The relative energy of the hydroxylation MECP is 1.3 kcal/mol lower than that of the dehydrogenation MECP, which accounts for cyclohexanol as the observed major product.

NONCOVALENT INTERACTIONS IN CONTROL OF THE SELECTIVITY Reactions with low activation barriers are expected to have low selectivity. Another currently accepted idea is the key role of the steric hindrance in controlling selectivity.29 A consequence is that, in general, the diastereoselectivity ratio, dr, is enhanced by the presence of bulky groups. For this reason, the discovery that reduction of artemisinic acid by diimide, HNNH, gives a high dr of 97:3 (Figure 6) appeared as a remarkable paradox since no factor justifies the preferential addition of the two hydrogens to

■

E


Article


Figure 7. Noncovalent stabilizing interactions (blue stronger than green), responsible for selectivity.

mechanism, in which proton followed by hydride are transferred to quinoline. It is thus essential that H2 remains in the coordination sphere of the metal and not be replaced by a stronger coordinating substrate. Consequently, the parent unsubstituted quinoline, which could coordinate to Ir, was detrimental to the reaction, but the uncoordinated bulky substituted quinoline was not.34

■

THE DEACTIVATION PATHWAYS AND THE ROBUSTNESS OF THE CATALYSTS Many computations focus on the formation of the desired products, ignoring catalyst deactivation and side-product formation.35 For instance, in alkene metathesis with the Schrock alkylidene complexes, a deactivation pathway initiated at the metallacyclobutane intermediate was identified. Computations showed that the olefin metathesis pathway goes via a trigonal bipyramidal (TBP) metallacyclobutane but that the square pyramidal (SP) isomer is at the origin of side-product formation and catalyst deactivation. In this latter isomer, the metal has an empty coordination site, which destroys the metallacyclobutane by β-C−H migration.36 The strategy to improve the catalytic process is to flatten the Gibbs energy profile for the productive pathway while disfavoring the deactivation pathway. DFT calculations revealed that the best catalyst is that which disfavors most the deactivation even though it also moderately disfavors the productive pathway (Figure 10).37

Figure 8. Molecular recognition model for the Mn-catalyzed C−H oxidation of iboprufen, including catalyst core (black), unrecognized substrate (red), and Kemp’s triacid recognition center (green) bound to reacting (blue) and protecting (orange) substrates.

contrast, the nonselective reaction of free ibuprofen has an energy barrier of ca. 20 kcal/mol, due to the lack of the double Hbond.

■

INNER- VERSUS OUTER-SPHERE CONTROL Preventing close contact between reactant and catalyst can also serve selectivity with the paradoxical result that the most bulky ligand is the one to react as shown in Figure 9 where αmethylquinoline is hydrogenated by [Ir(NHC)(PPh3)2(cod)] (NHC = substituted N-heterocyclic carbene, cod = 1,5cyclooctadiene). In this reaction, coordinated H2 was shown to be the hydrogenation reagent through an outer-sphere

Figure 10. Productive and unproductive pathways in alkene metathesis.

Activity is a key feature in catalysis, since it allows for mild reaction conditions. Nonetheless, in several applications, including energy conversion, robustness is equally relevant due to the high cost of most catalysts, which, once deactivated, may not be easily recycled. The potential harmful effects of some transition metals and the high standards required for product purity in many processes are also sensitive to robustness. From a conceptual point of view, this property is closely related to

Figure 9. IrIII catalyzed outer-sphere hydrogenation of quinoline. F


Article


of the substrate, one could ask whether face exchange is permitted during the chain walking. One can also be intrigued by selectivity of cyclopropane ring opening. The chain walking is promoted by allylic−CH activation to yield a Zr−hydride−allyl intermediate (Scheme 2). No energetically accessible alternative mechanism could be identified for this reaction. In this reaction σ- (or η1-) coordinated allyl intermediates are formed. In these intermediates the Zr−C rotation accounts for the chain walking along the same face of the carbon backbone. In the same complex, the σ-C−C rotation within the allyl motif accounts for E to Z π-bond isomerization. Finally, allylic C−H activation at a σ-cis conformation accounts for a switch in πbonding face. The free energy profile for these elementary steps reveals that allylic C−H activation, with ΔG# of 6−11 kcal/mol, controls the reaction. Consequently, all these processes are at work in the reaction conditions and the four isomers (A−D, Figure 13) of zirconocene β-ene-cyclopropane are produced in equilibrium via a common zirconocene-allyl-hydride intermediate I. In intermediates C and D, zirconocene is trans-periplanar to the less substituted cyclopropane C−C bond, and the large distortion required for cyclopropane ring opening kinetically prevents the reaction from proceeding (ΔG# > 35 kcal/mol) (Figure 13). In intermediates A and B, the C−C bond to be activated is syn-periplanar to the zirconocene moiety; as a result, the energy barrier drops by 30 kcal/mol, that is, ΔG# = 2−6 kcal/ mol for TSB‑IMaj and TSA‑Imin. The formation of the allyl−alkyl complex IMaj, which accounts for the major final diastereoisomer, is thermodynamically and kinetically favored over Imin, which accounts for the minor final diastereoisomer by a few kcal/mol. Though overestimated, this difference qualitatively accounts for the selectivity observed at low temperature. In order to rationalize the increase of selectivity observed upon heating, all possible pathways connecting Imin to IMaj were considered. The most favorable route is via intermediate I and is thus based on the reversible formation of Imin from I. In this epimerization reaction, the highest barrier relative to Imin is at 21 kcal/mol, that is 8 kcal/mol higher than the highest barrier computed along the reactants to products pathways. The overall reaction scheme is representative of a case of dynamic kinetic resolution, which rationalizes the increase of stereocontrol observed upon heating.41 As a result, this remote functionalization reaction profits from the unselective C−H activation that enables random chain walking and the reversibility of the cyclopropane ring opening, both on a floppy hilly surface that ensures microreversibility of all events until the system fully populates the lowest energy minima, here IMaj. More generally, in asymmetric synthesis, reactions that follow the dynamic kinetic resolution scheme require special care because of the risk to have an erroneous image of the factors that contribute to the formation of the preferred isomer by omitting step or conformer or by having an incomplete depiction of the intermediates that are in equilibrium.

selectivity. Catalysts should activate reactants but avoid selfdegradation.35 This seldom explored feature was studied on a collection of catalysts based on different transition metals and chelating P,N-donor ligands (Figure 11).38,39

Figure 11. Complexes considered for the study of degradation pathways initiated by C−H oxidation.

■

SELECTIVITY FROM REVERSIBILITY AND UNSELECTIVE EVENTS The counterintuitive increase of stereocontrol upon heating led our interest to the remote functionalization of chiral enecyclopropane derivatives with a high degree of stereocontrol.40 The reaction forms a quaternary asymmetric carbon with a 3:1 dr and high E/Z ratio of over 99:1 when the reaction is performed at room temperature with an overall yield of more than 50% (Figure 12). The reaction, assisted by zirconocene, involves double bond

Figure 12. Stereoselective remote bifunctionalization of ene-cyclopropane derivatives.

migration via zirconocene chain walking, selective cyclopropane ring opening, and two successive quenches by distinct electrophiles. The diastereocontrol reaches 98:2 when the reaction is heated up to 55 °C for 3 h before quenching. The high stereocontrol of the reaction is surprising relative to the number of steps and isomeric forms involved in the reaction. Since the coordination of Cp2Zr can occur on two faces of the double bond Scheme 2. Chain Walking

G


Article


Figure 13. Ene-cyclopropane remote functionalization (ΔG in kcal/mol).

■

CONCLUSIONS Is determining mechanisms by computation worth the effort? We believe so since optimizing a reaction is best achieved for a reaction whose mechanistic aspects are understood. More importantly, while the essential aspects of a bond transformation are overall understood, the great diversity of situations that are found in chemistry can set the stage for surprising divergences from standard patterns, justifying a detailed study in specific cases. An inquest is carried out af ter an unexplained case. We and other researchers, fascinated by the myriad of chemical mysteries that come to our attention, are always ready to embark on a new adventure. Our tools, methods and computers, are more powerful everyday, but the problems we tackle are more complex and the questions we ask are more probing. This keeps us all on our toes.

■

Eric Clot received his Ph.D. in 1995 from Université Paris-Sud at Orsay (Odile Eisenstein and Claude Leforestier directors) and joined the CNRS in 1996. Presently Research Director, his research focuses on the interplay between experiments and theory in multistep organic reactions catalyzed by transition metal complexes. Odile Eisenstein received her Ph.D. in 1977 from Université Paris-Sud at Orsay (Nguyen Trong Anh and L. Salem directors). Presently CNRS Emeritus Distinguish Research Director and member of the French Academy of Science, she keeps her interest in maintaining the dialog with experimental chemists. Ainara Nova received her Ph.D. in 2008 from Universitat Autònoma de Barcelona (A. Lledós and G. Ujaque directors). Postdoctoral associate at the Center of Excellence for Theoretical and Computational Chemistry at University of Oslo, she was awarded a RCN Young Talent project for computational design of catalytic CO2 functionalization. Lionel Perrin received his Ph.D. in 2004 from Université Montpellier (Odile Eisenstein and Laurent Maron directors) and joined the CNRS in 2005. Since 2013, he is group leader of Interface Theory Experiment: Mechanism and Modeling (ITEMM) at Université Lyon 1, he focuses on the interplay between experiments and theory for polymerization and multistep organic reactions catalyzed by organometallic species.

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Author Contributions

■

The manuscript was written through contributions of all authors. Funding

REFERENCES

(1) Perrin, L.; Carr, K. J. T.; McKay, D.; Mc Mullin, C. L.; Macgregor, S. A.; Eisenstein, O. Modeling and Rationalizing Organometallic Chemistry with Computation: where are We? Struct. Bonding (Berlin, Ger.) 2015, 176, 1−37. (2) Harvey, J. N. Ab initio transition state theory for polar reactions in solution. Faraday Discuss. 2010, 145, 487−505. (3) Plata, R. E.; Singleton, D. A. A case study of the mechanism of alcohol-mediated Morita-Baylis-Hillman reactions. The importance of experimental observations. J. Am. Chem. Soc. 2015, 137, 3811−3826. (4) Sperger, T.; Sanhueza, I. A.; Kalvet, I.; Schoenebeck, F. Computational Studies of Synthetically Relevant Homogeneous Organometallic Catalysis Involving Ni, Pd, Ir and Rh: An Overview of commonly Employed DFT Methods and Mechanistic Insights. Chem. Rev. 2015, 115, 9532−9586. (5) Biswas, B.; Singleton, D. A. Controlling Selectivity by controlling the path of trajectories. J. Am. Chem. Soc. 2015, 137, 14244−14248. (6) Jazzar, R.; Hitce, J.; Renaudat, A.; Sofack-Kreutzer, J.; Baudoin, O. Functionalization of Organic Molecules by Transition-Metal-Catalyzed C(sp3)−H Activation. Chem. - Eur. J. 2010, 16, 2654−2672. (7) Baudoin, O. Transition metal-catalyzed arylation of unactivated C(sp3)-H bonds. Chem. Soc. Rev. 2011, 40, 4902−4911.

E.C., O.E. and L.P. thank the CNRS, the Université de Montpellier, and the Université Lyon-1 for funding. D.B. and A.N. thank the Research Council of Norway for funding provided through the Centre of Excellence for Theoretical and Computational Chemistry (CTCC; Grant 179568/V30) and for a stipend to A.N. (Grant 221801/F20). D.B. thanks the EU REA for a Marie Curie Fellowship (Grant CompuWOC/618303). Notes

The authors declare no competing financial interest. Biographies David Balcells received his Ph.D. in 2006 from Universitat Autònoma de Barcelona and Institut Català d’Investigació Quimica, ICIQ, in Tarragona (G. Ujaque and F. Maseras directors). Researcher at the Center of Excellence for Theoretical and Computational Chemistry at University of Oslo with a Marie-Curie Fellowship, his research focuses on computational homogeneous catalysis by considering both on- and off-cycle reactions. H


Article


(26) (a) Shaik, S.; Hirao, H.; Kumar, D. Reactivity of High-Valent IronOxo Species in Enzymes and Synthetic Reagents: A Tale of Many States. Acc. Chem. Res. 2007, 40, 532−542. (b) Schroder, D.; Shaik, S.; Schwarz, H. Two-State Reactivity as a New Concept in Organometallic Chemistry. Acc. Chem. Res. 2000, 33, 139−145. (27) Conde, A.; Vilella, L.; Balcells, D.; Díaz-Requejo, M. M.; Lledós, A.; Pérez, P. J. Introducing Copper as Catalyst for Oxidative Alkane Dehydrogenation. J. Am. Chem. Soc. 2013, 135, 3887−3896. (28) Caballero, A.; Despagnet-Ayoub, E.; Diaz-Requejo, M. M.; DiazRodriguez, A.; Gonzalez-Nunez, M. E.; Mello, R.; Munoz, B. K.; Ojo, W. S.; Asensio, G.; Etienne, M.; Pérez, P. J. Silver-Catalyzed C-C Bond Formation Between Methane and Ethyl Diazoacetate in Supercritical CO2. Science 2011, 332, 835−838. (29) Kaselj, M.; Chung, W.-S.; le Noble, W. J. Face Selection in Addition and Elimination in Sterically Unbiased Systems. Chem. Rev. 1999, 99, 1387−1413. (30) Castro, B.; Chaudret, R.; Ricci, G.; Kurz, M.; Ochsenbein, P.; Kretzschmar, G.; Kraft, V.; Rossen, K.; Eisenstein, O. Nonclassical CH−π Supramolecular Interactions in Artemisinic Acid Favor a Single Conformation, Yielding High Diastereoselectivity in the Reduction with Diazene. J. Org. Chem. 2014, 79, 5939−5947. (31) (a) Di Cera, E. Site-Specific Thermodynamics: Understanding Cooperativity in Molecular Recognition. Chem. Rev. 1998, 98, 1563− 1591. (b) Kruppa, M.; Konig, B. Reversible Coordinative Bonds in Molecular Recognition. Chem. Rev. 2006, 106, 3520−3560. (c) Raynal, M.; Ballester, P.; Vidal-Ferran, A.; van Leeuwen, P. W. N. M. Supramolecular Catalysis. Part 1: Non-Covalent Interactions as a Tool for Building and Modifying Homogeneous Catalysts. Chem. Soc. Rev. 2014, 43, 1660−1733. (32) Das, S.; Incarvito, C. D.; Crabtree, R. H.; Brudvig, G. W. Molecular Recognition in the Selective Oxygenation of Saturated C-H Bonds by a Dimanganese Catalyst. Science 2006, 312, 1941−1943. (33) Balcells, D.; Moles, P.; Blakemore, J. D.; Raynaud, C.; Brudvig, G. W.; Crabtree, R. H.; Eisenstein, O. Molecular Recognition in Mncatalyzed C-H Oxidation. Reaction Mechanism and Origin of Selectivity from a DFT Perspective. Dalton Trans. 2009, 30, 5989−6000. (34) Dobereiner, G. E.; Nova, A.; Schley, N. D.; Hazari, N.; Miller, S. J.; Eisenstein, O.; Crabtree, R. H. Iridium-Catalyzed Hydrogenation of NHeterocyclic Compounds under Mild Conditions by an Outer-Sphere Pathway. J. Am. Chem. Soc. 2011, 133, 7547−7562. (35) Crabtree, R. H. Deactivation in Homogeneous Transition Metal Catalysis: Causes, Avoidance, and Cure. Chem. Rev. 2015, 115, 127− 150. (36) Leduc, A.-M.; Salameh, A.; Soulivong, D.; Chabanas, M.; Basset, J.-M.; Copéret, C.; Solans-Monfort, X.; Clot, E.; Eisenstein, O.; Böhm, V. P. W.; Röper, M. B-H transfer from the metallacyclobutane: a Key step in the deactivation and Byproduct formation for the Well defined Silica-Supported Rhenium alkylidene, alkene Metathesis Catalyst. J. Am. Chem. Soc. 2008, 130, 6288−6297. (37) Solans-Monfort, X.; Copéret, C.; Eisenstein, O. Shutting down Secondary Reaction Pathways: The Essential role of the Pyrrolyl Ligand in improving silica supported d0 ML4 Alkene metathesis catalyst from DFT Calculations. J. Am. Chem. Soc. 2010, 132, 7750−7757. (38) Nova, A.; Balcells, D. Does the Metal Protect the Ancillary Ligands? C-H Strengthening and Deactivation in Amines and Phosphines upon Metal-Binding. Chem. Commun. 2014, 50, 614−616. (39) Salamone, M.; Bietti, M. Tuning Reactivity and Selectivity in Hydrogen Atom Transfer from Aliphatic C-H Bonds to Alkoxyl Radicals: Role of Structural and Medium Effects. Acc. Chem. Res. 2015, 48, 2895−2903. (40) Vasseur, A.; Perrin, L.; Eisenstein, O.; Marek, I. Remote functionalization of hydrocarbons with reversibility enhanced stereocontrol. Chem. Sci. 2015, 6, 2770−2776. (41) Faber, K. Non-Sequential Processes for the Transformation of a Racemate into a Single Stereoisomeric Product: Proposal for Stereochemical Classification. Chem. - Eur. J. 2001, 7, 5004−5010.

(8) Murai, S.; Kakiuchi, F.; Sekine, S.; Tanaka, Y.; Kamatani, A.; Sonoda, M.; Chatani, N. Efficient catalytic addition of aromatic carbonhydrogen bonds to olefins. Nature 1993, 366, 529−531. (9) Campeau, L.-C.; Parisien, M.; Leblanc, M.; Fagnou, K. Biaryl Synthesis via Direct Arylation: Establishment of an Efficient Catalyst for Intramolecular Processes. J. Am. Chem. Soc. 2004, 126, 9186−9187. (10) Balcells, D.; Clot, E.; Eisenstein, O. C−H Bond Activation in Transition Metal Species from a Computational Perspective. Chem. Rev. 2010, 110, 749−823. (11) Baudoin, O.; Herrbach, A.; Guéritte, F. The Palladium-Catalyzed C−H Activation of Benzylic gem-Dialkyl Groups. Angew. Chem., Int. Ed. 2003, 42, 5736−5740. (12) Chaumontet, M.; Piccardi, R.; Audic, N.; Hitce, J.; Peglion, J.-L.; Clot, E.; Baudoin, O. Synthesis of Benzocyclobutenes by PalladiumCatalyzed C-H Activation of Methyl Groups: Method and Mechanistic Study. J. Am. Chem. Soc. 2008, 130, 15157−15166. (13) Clot, E.; Eisenstein, O.; Jasim, N.; Macgregor, S. A.; McGrady, J. E.; Perutz, R. N. C−F and C−H bond activation of fluorobenzenes and fluoropyridines at transition metal centers: how fluorine tips the scales. Acc. Chem. Res. 2011, 44, 333−348. (14) Kefalidis, C. E.; Baudoin, O.; Clot, E. DFT study of the mechanism of benzocyclobutene formation by palladium-catalysed C(sp3)-H activation: role of the nature of the base and the phosphine. Dalton Trans. 2010, 39, 10528−10535. (15) Rousseaux, S.; Davi, M.; Sofack-Kreutzer, J.; Pierre, C.; Kefalidis, C. E.; Clot, E.; Fagnou, K.; Baudoin, O. Intramolecular PalladiumCatalyzed Alkane C-H Arylation from Aryl Chlorides. J. Am. Chem. Soc. 2010, 132, 10706−10716. (16) Holstein, P. M.; Vogler, M.; Larini, P.; Pilet, G.; Clot, E.; Baudoin, O. Efficient Pd-0-Catalyzed Asymmetric Activation of Primary and Secondary C-H Bonds Enabled by Modular Binepine Ligands and Carbonate Bases. ACS Catal. 2015, 5, 4300−4308. (17) Kefalidis, C. E.; Davi, M.; Holstein, P. M.; Clot, E.; Baudoin, O. Mechanistic Study of the Selectivity of Olefin versus Cyclobutene Formation by Palladium(0)-Catalyzed Intramolecular C(sp3)-H Activation. J. Org. Chem. 2014, 79, 11903−11910. (18) Groves, J. T.; McClusky, G. A. Aliphatic Hydroxylation via Oxygen Rebound- Oxygen-Transfer Catalyzed by Iron. J. Am. Chem. Soc. 1976, 98, 859−861. (19) Kim, C.; Dong, Y. H.; Que, L. Modeling Nonheme Diiron Enzymes: Hydrocarbon Hydroxylation and Desaturation by a High Valent Fe2O2 Diamond Core. J. Am. Chem. Soc. 1997, 119, 3635−3636. (20) (a) Meunier, B.; de Visser, S. P.; Shaik, S. Mechanism of Oxidation Reactions Catalyzed by Cytochrome P450 Enzymes. Chem. Rev. 2004, 104, 3947−3980. (b) Shaik, S.; Kumar, D.; de Visser, S. P.; Altun, A.; Thiel, W. Theoretical Perspective on the Structure and Mechanism of Cytochrome P450 Enzymes. Chem. Rev. 2005, 105, 2279−2328. (c) Ortiz de Montellano, P. R. Hydrocarbon Hydroxylation by Cytochrome P450 Enzymes. Chem. Rev. 2010, 110, 932−948. (21) Fox, B. G.; Lyle, K. S.; Rogge, C. E. Reactions of the Diiron Enzyme Stearoyl-Acyl Carrier Protein Desaturase. Acc. Chem. Res. 2004, 37, 421−429. (22) Hull, J. F.; Balcells, D.; Sauer, E. L. O.; Raynaud, C.; Brudvig, G. W.; Crabtree, R. H.; Eisenstein, O. Manganese Catalysts for C-H Activation: An Experimental/Theoretical Study Identifies the Stereoelectronic Factor that Controls the Switch between Hydroxylation and Desaturation Pathways. J. Am. Chem. Soc. 2010, 132, 7605−7616. (23) Balcells, D.; Raynaud, C.; Crabtree, R. H.; Eisenstein, O. The Rebound Mechanism in Catalytic C-H Oxidation by Mn(O) (tpp) (Cl) from DFT Studies: Electronic Nature of the Active Species. Chem. Commun. 2008, 744−746. (24) Khenkin, A. M.; Kumar, D.; Shaik, S.; Neumann, R. Characterization of Manganese(V)-Oxo Polyoxometalate Intermediates and their Properties in Oxygen-Transfer Reactions. J. Am. Chem. Soc. 2006, 128, 15451−15460. (25) Balcells, D.; Raynaud, C.; Crabtree, R. H.; Eisenstein, O. A Rational Basis for the Axial Ligand Effect in C-H Oxidation by [MnO(porphyrin)(X)]+ (X = H2O, OH−, O2‑) from a DFT Study. Inorg. Chem. 2008, 47, 10090−10099. I


Deciphering Selectivity in Organic Reactions: A Multifaceted Problem

Recommend Documents